Unify GraphImpl RDDs + other graph load optimizations

This PR makes the following changes, primarily in e4fbd32:

1. *Unify RDDs to avoid zipPartitions.* A graph used to be four RDDs: vertices, edges, routing table, and triplet view. This commit merges them down to two: vertices (with routing table), and edges (with replicated vertices).

2. *Avoid duplicate shuffle in graph building.* We used to do two shuffles when building a graph: one to extract routing information from the edges and move it to the vertices, and another to find nonexistent vertices referred to by edges. With this commit, the latter is done as a side effect of the former.

3. *Avoid no-op shuffle when joins are fully eliminated.* This is a side effect of unifying the edges and the triplet view.

4. *Join elimination for mapTriplets.*

5. *Ship only the needed vertex attributes when upgrading the triplet view.* If the triplet view already contains source attributes, and we now need both attributes, only ship destination attributes rather than re-shipping both. This is done in `ReplicatedVertexView#upgrade`.

Author: Ankur Dave <[email protected]>

Closes #497 from ankurdave/unify-rdds and squashes the following commits:

332ab43 [Ankur Dave] Merge remote-tracking branch 'apache-spark/master' into unify-rdds
4933e2e [Ankur Dave] Exclude RoutingTable from binary compatibility check
5ba8789 [Ankur Dave] Add GraphX upgrade guide from Spark 0.9.1
13ac845 [Ankur Dave] Merge remote-tracking branch 'apache-spark/master' into unify-rdds
a04765c [Ankur Dave] Remove unnecessary toOps call
57202e8 [Ankur Dave] Replace case with pair parameter
75af062 [Ankur Dave] Add explicit return types
04d3ae5 [Ankur Dave] Convert implicit parameter to context bound
c88b269 [Ankur Dave] Revert upgradeIterator to if-in-a-loop
0d3584c [Ankur Dave] EdgePartition.size should be val
2a928b2 [Ankur Dave] Set locality wait
10b3596 [Ankur Dave] Clean up public API
ae36110 [Ankur Dave] Fix style errors
e4fbd32 [Ankur Dave] Unify GraphImpl RDDs + other graph load optimizations
d6d60e2 [Ankur Dave] In GraphLoader, coalesce to minEdgePartitions
62c7b78 [Ankur Dave] In Analytics, take PageRank numIter
d64e8d4 [Ankur Dave] Log current Pregel iteration

Author: ankurdave

docs/graphx-programming-guide.md

@@ -86,6 +86,12 @@ support the [Bagel API](api/scala/index.html#org.apache.spark.bagel.package) and
 [Bagel programming guide](bagel-programming-guide.html). However, we encourage Bagel users to
 explore the new GraphX API and comment on issues that may complicate the transition from Bagel.
 
+## Upgrade Guide from Spark 0.9.1
+
+GraphX in Spark {{site.SPARK_VERSION}} contains one user-facing interface change from Spark 0.9.1. [`EdgeRDD`][EdgeRDD] may now store adjacent vertex attributes to construct the triplets, so it has gained a type parameter. The edges of a graph of type `Graph[VD, ED]` are of type `EdgeRDD[ED, VD]` rather than `EdgeRDD[ED]`.
+
+[EdgeRDD]: api/scala/index.html#org.apache.spark.graphx.EdgeRDD
+
 # Getting Started
 
 To get started you first need to import Spark and GraphX into your project, as follows:
@@ -145,12 +151,12 @@ the vertices and edges of the graph:
 {% highlight scala %}
 class Graph[VD, ED] {
   val vertices: VertexRDD[VD]
-  val edges: EdgeRDD[ED]
+  val edges: EdgeRDD[ED, VD]
 }
 {% endhighlight %}
 
-The classes `VertexRDD[VD]` and `EdgeRDD[ED]` extend and are optimized versions of `RDD[(VertexID,
-VD)]` and `RDD[Edge[ED]]` respectively.  Both `VertexRDD[VD]` and `EdgeRDD[ED]` provide  additional
+The classes `VertexRDD[VD]` and `EdgeRDD[ED, VD]` extend and are optimized versions of `RDD[(VertexID,
+VD)]` and `RDD[Edge[ED]]` respectively.  Both `VertexRDD[VD]` and `EdgeRDD[ED, VD]` provide  additional
 functionality built around graph computation and leverage internal optimizations.  We discuss the
 `VertexRDD` and `EdgeRDD` API in greater detail in the section on [vertex and edge
 RDDs](#vertex_and_edge_rdds) but for now they can be thought of as simply RDDs of the form:
@@ -302,7 +308,7 @@ class Graph[VD, ED] {
   val degrees: VertexRDD[Int]
   // Views of the graph as collections =============================================================
   val vertices: VertexRDD[VD]
-  val edges: EdgeRDD[ED]
+  val edges: EdgeRDD[ED, VD]
   val triplets: RDD[EdgeTriplet[VD, ED]]
   // Functions for caching graphs ==================================================================
   def persist(newLevel: StorageLevel = StorageLevel.MEMORY_ONLY): Graph[VD, ED]
@@ -908,7 +914,7 @@ val setC: VertexRDD[Double] = setA.innerJoin(setB)((id, a, b) => a + b)
 
 ## EdgeRDDs
 
-The `EdgeRDD[ED]`, which extends `RDD[Edge[ED]]` organizes the edges in blocks partitioned using one
+The `EdgeRDD[ED, VD]`, which extends `RDD[Edge[ED]]` organizes the edges in blocks partitioned using one
 of the various partitioning strategies defined in [`PartitionStrategy`][PartitionStrategy].  Within
 each partition, edge attributes and adjacency structure, are stored separately enabling maximum
 reuse when changing attribute values.
@@ -918,11 +924,11 @@ reuse when changing attribute values.
 The three additional functions exposed by the `EdgeRDD` are:
 {% highlight scala %}
 // Transform the edge attributes while preserving the structure
-def mapValues[ED2](f: Edge[ED] => ED2): EdgeRDD[ED2]
+def mapValues[ED2](f: Edge[ED] => ED2): EdgeRDD[ED2, VD]
 // Revere the edges reusing both attributes and structure
-def reverse: EdgeRDD[ED]
+def reverse: EdgeRDD[ED, VD]
 // Join two `EdgeRDD`s partitioned using the same partitioning strategy.
-def innerJoin[ED2, ED3](other: EdgeRDD[ED2])(f: (VertexId, VertexId, ED, ED2) => ED3): EdgeRDD[ED3]
+def innerJoin[ED2, ED3](other: EdgeRDD[ED2, VD])(f: (VertexId, VertexId, ED, ED2) => ED3): EdgeRDD[ED3, VD]
 {% endhighlight %}
 
 In most applications we have found that operations on the `EdgeRDD` are accomplished through the

graphx/src/main/scala/org/apache/spark/graphx/EdgeRDD.scala

@@ -20,16 +20,19 @@ package org.apache.spark.graphx
 import scala.reflect.{classTag, ClassTag}
 
 import org.apache.spark.{OneToOneDependency, Partition, Partitioner, TaskContext}
-import org.apache.spark.graphx.impl.EdgePartition
 import org.apache.spark.rdd.RDD
 import org.apache.spark.storage.StorageLevel
 
+import org.apache.spark.graphx.impl.EdgePartition
+
 /**
- * `EdgeRDD[ED]` extends `RDD[Edge[ED]]` by storing the edges in columnar format on each partition
- * for performance.
+ * `EdgeRDD[ED, VD]` extends `RDD[Edge[ED]]` by storing the edges in columnar format on each
+ * partition for performance. It may additionally store the vertex attributes associated with each
+ * edge to provide the triplet view. Shipping of the vertex attributes is managed by
+ * `impl.ReplicatedVertexView`.
  */
-class EdgeRDD[@specialized ED: ClassTag](
-    val partitionsRDD: RDD[(PartitionID, EdgePartition[ED])])
+class EdgeRDD[@specialized ED: ClassTag, VD: ClassTag](
+    val partitionsRDD: RDD[(PartitionID, EdgePartition[ED, VD])])
   extends RDD[Edge[ED]](partitionsRDD.context, List(new OneToOneDependency(partitionsRDD))) {
 
   partitionsRDD.setName("EdgeRDD")
@@ -45,8 +48,12 @@ class EdgeRDD[@specialized ED: ClassTag](
     partitionsRDD.partitioner.orElse(Some(Partitioner.defaultPartitioner(partitionsRDD)))
 
   override def compute(part: Partition, context: TaskContext): Iterator[Edge[ED]] = {
-    val p = firstParent[(PartitionID, EdgePartition[ED])].iterator(part, context)
-    p.next._2.iterator.map(_.copy())
+    val p = firstParent[(PartitionID, EdgePartition[ED, VD])].iterator(part, context)
+    if (p.hasNext) {
+      p.next._2.iterator.map(_.copy())
+    } else {
+      Iterator.empty
+    }
   }
 
   override def collect(): Array[Edge[ED]] = this.map(_.copy()).collect()
@@ -61,11 +68,15 @@ class EdgeRDD[@specialized ED: ClassTag](
     this
   }
 
-  private[graphx] def mapEdgePartitions[ED2: ClassTag](
-      f: (PartitionID, EdgePartition[ED]) => EdgePartition[ED2]): EdgeRDD[ED2] = {
-    new EdgeRDD[ED2](partitionsRDD.mapPartitions({ iter =>
-      val (pid, ep) = iter.next()
-      Iterator(Tuple2(pid, f(pid, ep)))
+  private[graphx] def mapEdgePartitions[ED2: ClassTag, VD2: ClassTag](
+      f: (PartitionID, EdgePartition[ED, VD]) => EdgePartition[ED2, VD2]): EdgeRDD[ED2, VD2] = {
+    new EdgeRDD[ED2, VD2](partitionsRDD.mapPartitions({ iter =>
+      if (iter.hasNext) {
+        val (pid, ep) = iter.next()
+        Iterator(Tuple2(pid, f(pid, ep)))
+      } else {
+        Iterator.empty
+      }
     }, preservesPartitioning = true))
   }
 
@@ -76,15 +87,22 @@ class EdgeRDD[@specialized ED: ClassTag](
    * @param f the function from an edge to a new edge value
    * @return a new EdgeRDD containing the new edge values
    */
-  def mapValues[ED2: ClassTag](f: Edge[ED] => ED2): EdgeRDD[ED2] =
+  def mapValues[ED2: ClassTag](f: Edge[ED] => ED2): EdgeRDD[ED2, VD] =
     mapEdgePartitions((pid, part) => part.map(f))
 
   /**
    * Reverse all the edges in this RDD.
    *
    * @return a new EdgeRDD containing all the edges reversed
    */
-  def reverse: EdgeRDD[ED] = mapEdgePartitions((pid, part) => part.reverse)
+  def reverse: EdgeRDD[ED, VD] = mapEdgePartitions((pid, part) => part.reverse)
+
+  /** Removes all edges but those matching `epred` and where both vertices match `vpred`. */
+  def filter(
+      epred: EdgeTriplet[VD, ED] => Boolean,
+      vpred: (VertexId, VD) => Boolean): EdgeRDD[ED, VD] = {
+    mapEdgePartitions((pid, part) => part.filter(epred, vpred))
+  }
 
   /**
    * Inner joins this EdgeRDD with another EdgeRDD, assuming both are partitioned using the same
@@ -96,19 +114,15 @@ class EdgeRDD[@specialized ED: ClassTag](
    *         with values supplied by `f`
    */
   def innerJoin[ED2: ClassTag, ED3: ClassTag]
-      (other: EdgeRDD[ED2])
-      (f: (VertexId, VertexId, ED, ED2) => ED3): EdgeRDD[ED3] = {
+      (other: EdgeRDD[ED2, _])
+      (f: (VertexId, VertexId, ED, ED2) => ED3): EdgeRDD[ED3, VD] = {
     val ed2Tag = classTag[ED2]
     val ed3Tag = classTag[ED3]
-    new EdgeRDD[ED3](partitionsRDD.zipPartitions(other.partitionsRDD, true) {
+    new EdgeRDD[ED3, VD](partitionsRDD.zipPartitions(other.partitionsRDD, true) {
       (thisIter, otherIter) =>
         val (pid, thisEPart) = thisIter.next()
         val (_, otherEPart) = otherIter.next()
         Iterator(Tuple2(pid, thisEPart.innerJoin(otherEPart)(f)(ed2Tag, ed3Tag)))
     })
   }
-
-  private[graphx] def collectVertexIds(): RDD[VertexId] = {
-    partitionsRDD.flatMap { case (_, p) => Array.concat(p.srcIds, p.dstIds) }
-  }
 }

graphx/src/main/scala/org/apache/spark/graphx/EdgeTriplet.scala

@@ -63,4 +63,6 @@ class EdgeTriplet[VD, ED] extends Edge[ED] {
     if (srcId == vid) srcAttr else { assert(dstId == vid); dstAttr }
 
   override def toString = ((srcId, srcAttr), (dstId, dstAttr), attr).toString()
+
+  def toTuple: ((VertexId, VD), (VertexId, VD), ED) = ((srcId, srcAttr), (dstId, dstAttr), attr)
 }

graphx/src/main/scala/org/apache/spark/graphx/Graph.scala

@@ -59,7 +59,7 @@ abstract class Graph[VD: ClassTag, ED: ClassTag] protected () extends Serializab
    * along with their vertex data.
    *
    */
-  @transient val edges: EdgeRDD[ED]
+  @transient val edges: EdgeRDD[ED, VD]
 
   /**
    * An RDD containing the edge triplets, which are edges along with the vertex data associated with

graphx/src/main/scala/org/apache/spark/graphx/GraphKryoRegistrator.scala

@@ -19,10 +19,11 @@ package org.apache.spark.graphx
 
 import com.esotericsoftware.kryo.Kryo
 
-import org.apache.spark.graphx.impl._
 import org.apache.spark.serializer.KryoRegistrator
-import org.apache.spark.util.collection.BitSet
 import org.apache.spark.util.BoundedPriorityQueue
+import org.apache.spark.util.collection.BitSet
+
+import org.apache.spark.graphx.impl._
 
 /**
  * Registers GraphX classes with Kryo for improved performance.
@@ -33,8 +34,9 @@ class GraphKryoRegistrator extends KryoRegistrator {
     kryo.register(classOf[Edge[Object]])
     kryo.register(classOf[MessageToPartition[Object]])
     kryo.register(classOf[VertexBroadcastMsg[Object]])
+    kryo.register(classOf[RoutingTableMessage])
     kryo.register(classOf[(VertexId, Object)])
-    kryo.register(classOf[EdgePartition[Object]])
+    kryo.register(classOf[EdgePartition[Object, Object]])
     kryo.register(classOf[BitSet])
     kryo.register(classOf[VertexIdToIndexMap])
     kryo.register(classOf[VertexAttributeBlock[Object]])

graphx/src/main/scala/org/apache/spark/graphx/GraphLoader.scala

@@ -47,8 +47,7 @@ object GraphLoader extends Logging {
    * @param path the path to the file (e.g., /home/data/file or hdfs://file)
    * @param canonicalOrientation whether to orient edges in the positive
    *        direction
-   * @param minEdgePartitions the number of partitions for the
-   *        the edge RDD
+   * @param minEdgePartitions the number of partitions for the edge RDD
    */
   def edgeListFile(
       sc: SparkContext,
@@ -60,8 +59,9 @@ object GraphLoader extends Logging {
     val startTime = System.currentTimeMillis
 
     // Parse the edge data table directly into edge partitions
-    val edges = sc.textFile(path, minEdgePartitions).mapPartitionsWithIndex { (pid, iter) =>
-      val builder = new EdgePartitionBuilder[Int]
+    val lines = sc.textFile(path, minEdgePartitions).coalesce(minEdgePartitions)
+    val edges = lines.mapPartitionsWithIndex { (pid, iter) =>
+      val builder = new EdgePartitionBuilder[Int, Int]
       iter.foreach { line =>
         if (!line.isEmpty && line(0) != '#') {
           val lineArray = line.split("\\s+")
@@ -78,7 +78,7 @@ object GraphLoader extends Logging {
         }
       }
       Iterator((pid, builder.toEdgePartition))
-    }.cache()
+    }.cache().setName("GraphLoader.edgeListFile - edges (%s)".format(path))
     edges.count()
 
     logInfo("It took %d ms to load the edges".format(System.currentTimeMillis - startTime))

graphx/src/main/scala/org/apache/spark/graphx/GraphOps.scala

@@ -18,11 +18,13 @@
 package org.apache.spark.graphx
 
 import scala.reflect.ClassTag
-import org.apache.spark.SparkContext._
+import scala.util.Random
+
 import org.apache.spark.SparkException
-import org.apache.spark.graphx.lib._
+import org.apache.spark.SparkContext._
 import org.apache.spark.rdd.RDD
-import scala.util.Random
+
+import org.apache.spark.graphx.lib._
 
 /**
  * Contains additional functionality for [[Graph]]. All operations are expressed in terms of the
@@ -43,19 +45,22 @@ class GraphOps[VD: ClassTag, ED: ClassTag](graph: Graph[VD, ED]) extends Seriali
    * The in-degree of each vertex in the graph.
    * @note Vertices with no in-edges are not returned in the resulting RDD.
    */
-  @transient lazy val inDegrees: VertexRDD[Int] = degreesRDD(EdgeDirection.In)
+  @transient lazy val inDegrees: VertexRDD[Int] =
+    degreesRDD(EdgeDirection.In).setName("GraphOps.inDegrees")
 
   /**
    * The out-degree of each vertex in the graph.
    * @note Vertices with no out-edges are not returned in the resulting RDD.
    */
-  @transient lazy val outDegrees: VertexRDD[Int] = degreesRDD(EdgeDirection.Out)
+  @transient lazy val outDegrees: VertexRDD[Int] =
+    degreesRDD(EdgeDirection.Out).setName("GraphOps.outDegrees")
 
   /**
    * The degree of each vertex in the graph.
    * @note Vertices with no edges are not returned in the resulting RDD.
    */
-  @transient lazy val degrees: VertexRDD[Int] = degreesRDD(EdgeDirection.Either)
+  @transient lazy val degrees: VertexRDD[Int] =
+    degreesRDD(EdgeDirection.Either).setName("GraphOps.degrees")
 
   /**
    * Computes the neighboring vertex degrees.

graphx/src/main/scala/org/apache/spark/graphx/Pregel.scala

@@ -18,6 +18,7 @@
 package org.apache.spark.graphx
 
 import scala.reflect.ClassTag
+import org.apache.spark.Logging
 
 
 /**
@@ -52,7 +53,7 @@ import scala.reflect.ClassTag
  * }}}
  *
  */
-object Pregel {
+object Pregel extends Logging {
 
   /**
    * Execute a Pregel-like iterative vertex-parallel abstraction.  The
@@ -142,6 +143,9 @@ object Pregel {
       // hides oldMessages (depended on by newVerts), newVerts (depended on by messages), and the
       // vertices of prevG (depended on by newVerts, oldMessages, and the vertices of g).
       activeMessages = messages.count()
+
+      logInfo("Pregel finished iteration " + i)
+
       // Unpersist the RDDs hidden by newly-materialized RDDs
       oldMessages.unpersist(blocking=false)
       newVerts.unpersist(blocking=false)